Longitudinal Methods (Cohort Analysis, Life Tables)

Abstract

This article described the origins of longitudinal methods and the variety of longitudinal models applied in geographic research. Longitudinal models frequently are termed hazard models, duration models, or event-history analysis. Early precursors to event-history analysis are studies of survival based on life tables, and cohort analysis in life course research. Longitudinal methods are applicable for the study of temporal and spatial processes, although the former are developed more extensively. This article explains important terminology used in longitudinal modeling and numerous theoretical time dependencies. Within life course research, longitudinal methods are used to understand multiple time dependencies. These are frequently referred to as parallel events across various careers and trajectories. Next, a variety of longitudinal models are described representing parametric, semiparametric, and nonparametric approaches applied in continuous-time and discrete-time dimensions. These include parametric continuous-time hazard models, Cox's proportional hazards model, discrete-time hazard models, log-rate models, and event-history diffusion analysis that extends longitudinal models to spatial processes. The final section addresses current and future developments in longitudinal modeling in geography.